HYPOTHESIS: Central nervous system dysfunction is a central pathogenic mechanism in the CFS spectrum of illnesses. Cerebrospinal fluid provides a "window" into potential dysfunctional regulatory, innate immune, and neurological pathways. Neurons, glial cells, epithelial choroid plexus and leptomeningeal cells may be sources of CFS-related proteins. Despite the diverse clinical syndromes, the CFS-related proteome is the same, suggesting a unified pathogenesis. DATA: We have performed tandem mass spectrometry (MS-MS) on cerebrospinal fluid from CFS and healthy control subjects. Traditional and support vector machine (SVM) learning statistical analyses identified nearly identical CFS-related proteomes. A specific pattern of proteins (biosignature) predicted CFS with a significant odds ratio of 34.5 and concordance of 80%. Amyloidogenic proteins, antiproteases, Ig lambda, heme and Fe scavengers, and regulatory prohormones were associated with CFS. This is the first predictive model of CFS to be defined solely from objective data. PLAN: Recruit a new set of CFS and HC subjects (n=50 per group, "cohort 4") to a cross-sectional "training-test" study design. (A) Perform qualitative MS-MS to identify proteins in all samples of cohort 4. Train the SVM algorithm with cohort 4, then test the output "classifier" on an independent set of 42 samples (cohort 3). Determine the prediction accuracy, sensitivity and specificity of the SVM classifier. (B) Perform quantitative MS-MS on pooled CFS and pooled control samples by labeling one with O16 and the other with O18. Mix the samples and identify peptides (and their parent proteins) with O16/O18 ratios that are significantly higher or lower in CFS than controls. (C) These CFS-related proteins will be measured using novel, high sensitivity, luciferase- fusion protein competition immunoassays. Significant concentrations differences between CFS and control and between the 2 cohorts will define protein biomarkers and their sensitivity, specificity and predictive accuracy. (D) Subjective psychometric and other input variables will be tested by SVM learning to define a highly predictive model of CFS. The subjective results and objective proteomic results will also be analyzed to determine if the biomarkers are highly correlated with fatigue, systemic hyperalgesia, or other components of the CFS spectrum of illness. These methods and biomarkers may be of diagnostic value. They will be useful for assessing longitudinal changes in disease severity, phenotype, or the effects of treatment. [unreadable] [unreadable] [unreadable]